Steve Lohr had an article in the NYT promoting the possibility that generative intelligence from A.I. will bring us scientific breakthroughs quicker. A new startup called “Lila” is trying to achieve this. George Church of Harvard is onboard.
Back on Sun., Dec. 8, 2019, 60 Minutes on ran a very optimistic segment in which Church says that through his lab’s work on gene editing, age reversal for dogs “might be a couple years away and then that takes another ten years to get through the human clinical trials” (Church as quoted in Pelley 2019).
In Lohr’s recent article, Church is quoted as saying ““I think science is a really good topic for A.I.” (Church as quoted in Lohr, p. B5). The article describes science as basically a mechanical process of trial and error. Some science is like that, like when Gerhard Domagk had his lab crank through hundreds of chemicals to find one (Prontosil) that was a broad spectrum antibiotic. Maybe A.I. could more efficiently crank through a large set of possibilities. The only example of medical advance through A.I. in the article is that “Lila’s A.I. has generated novel antibodies to fight disease” (Lohr, p. B5).
A.I. can combine what is known in novel ways and produce text that is new, but is not necessarily sensible, correct, or useful, let alone a profound leap.
So it is not clear to me how well A.I. could help define and prioritize the possibilities. Lila scientists are feeding their A.I. program scientific literature, presumably weighting differing views by some bibliometric measures, like citations or journal rankings. But often a leap or breakthrough is at first rejected by the top journals, and not heavily cited by the establishment.
I do not see how A.I. could identify those early breakthroughs, much less be the source of them. And making and identifying such breakthroughs are key steps in scientific progress.
I was was pumped when I heard Church’s optimism in 2019 for longevity breakthroughs. But now it is more than five years later, and I have not seen claims of age reversal for dogs, let alone for humans. Maybe Covid delayed progress. Or maybe Church is not a good judge of what is required for scientific breakthroughs. This latter possibility seems more likely given Church’s hyper-enthusiasm for generative A.I.
Steve Lohr’s article is:
(Note: the online version of the Steve Lohr article has the date March 10, 2025, and has the title “The Quest for A.I. ‘Scientific Superintelligence’.”)
A transcript of the 60 Minutes segment on Church is:
Pelley, Scott. “A Harvard Geneticist’s Goal: To Protect Humans from Viruses, Genetic Diseases, and Aging.” In 60 Minutes. CBS News, (Sun., Dec. 8, 2019).